How We Boosted Nginx Performance 50× by Tuning Gzip Settings

This article records a practical Nginx tuning process that increased overall server performance by about 50 times through step‑by‑step analysis and implementation.

1. Requirement Background

The project needed to display data in multiple rounds, with an estimated peak of 90,000 QPS. As backend engineers we initially prepared the API, hierarchical caching, machine scaling, and maxed out threads, but frequent data updates forced us to generate static files and serve them via CDN.

The architecture is illustrated below:

After each data update, new static files are generated and CDN is refreshed, causing a massive surge of origin requests that could hold up to 90,000 QPS on the application servers.

2. First Load Test

Two data‑center rooms with 40 machines (4 C each) served 25 KB files at 50,000 QPS, pushing CPU to 90 %.

Adding more machines did not solve the bottleneck; CPU remained saturated even after doubling the servers to 80.

We suspected the enabled gzip compression might be consuming CPU.

server {
    listen 80;
    gzip on;
    gzip_disable "msie6";
    gzip_vary on;
    gzip_proxied any;
    gzip_comp_level 6;
    gzip_buffers 16 8k;
    gzip_http_version 1.1;
    gzip_types text/plain application/css text/css application/xml text/javascript application/javascript application/x-javascript;
    ...
}

3. Second Load Test

We reduced the gzip compression level from 6 to 2 to lower CPU load. gzip_comp_level 2; CPU still saturated quickly, but QPS barely reached 90,000, confirming gzip was the main CPU consumer.

4. Understanding Nginx Gzip

Nginx is renowned for high‑performance static serving, yet gzip compression can become a CPU bottleneck. Gzip works by removing redundant characters in static files (HTML, CSS, JS) to reduce bandwidth.

Gzip in Nginx has two modes:

Dynamic compression : The server compresses responses on the fly, consuming CPU for each request.

Static compression : Pre‑compressed .gz files are served directly; if a .gz file is missing, the original file is used.

Static compression requires the ngx_http_gzip_static_module, which is already compiled in the JDos Nginx distribution.

We generated .gz files locally using GZIPOutputStream and enabled static compression in Nginx:

gzip_static on;

5. Final Results

With static gzip, 40 machines handled the 90,000 QPS load using only 7 % CPU. Pushing further, CPU grew slowly while network output reached 89 MB/s; QPS climbed to 270,000.

Overall, QPS increased from 50 k to 270 k (≈5×), CPU usage dropped from 90 % to 7 % (≈10×), delivering more than a 50‑fold performance improvement.

6. Conclusion

Static gzip compression provides overwhelming benefits for immutable files, eliminating CPU and bandwidth waste. Dynamic compression remains suitable for API responses that change frequently. This hands‑on experience deepened our understanding of Nginx’s gzip mechanisms and highlighted a valuable optimization path for backend engineers.